Accuracy Evaluation of Transposed Convolution-Based Quantized Neural Networks